Soil moisture is an important part of global energy cycle. Obtaining quantitative soil moisture accurately is of great significance for environmental protection,agricultural production monitoring,global change study, etc. There are many remote sensing platforms providing soil moisture data in different temporal and spatial scales,each of which has its own advantages and disadvantages. By using AMSR-E soil moisture data gained by passive microwave radiometers and MODIS data,taking Mongolia and Asia as the verification zone,we developed a new downscaling algorithm for large-scale soil moisture data under the framework of the spectrum downscaling algorithm by improving the model fitting equation and introducing the evapotranspiration model of remote sensing physics. The comparison with ground measured data from the Asia-Australia Monsoon Project of Coordinated Energy and water-cycle Observation Project (CEOP) in Mongolia study area shows that the downscaled soil moisture data and its changing trends are in good agreement with measured data. Hence it is confirmed that the downscaling method is of high accuracy and credibility.